Improving HLA typing imputation accuracy and eplet identification with local next-generation sequencing training data.

Assessing donor/recipient HLA compatibility at the eplet level requires second field DNA typings but these are not always available. These can be estimated from lower-resolution data either manually or with computational tools currently relying, at best, on data containing typing ambiguities. We gathered NGS typing data from 61,393 individuals in 17 French laboratories, for loci A, B, and C (100% of typings), DRB1 and DQB1 (95.5%), DQA1 (39.6%), DRB3/4/5, DPB1, and DPA1 (10.5%). We developed HaploSFHI, a modified iterative maximum likelihood algorithm, to impute second field HLA typings from low- or intermediate-resolution ones. Compared with the reference tools HaploStats, HLA-EMMA, and HLA-Upgrade, HaploSFHI provided more accurate predictions across all loci on two French test sets and four European-independent test sets. Only HaploSFHI could impute DQA1, and solely HaploSFHI and HaploStats provided DRB3/4/5 imputations. The improved performance of HaploSFHI was due to our local and nonambiguous data. We provided explanations for the most common imputation errors and pinpointed the variability of a low number of low-resolution haplotypes. We thus provided guidance to select individuals for whom sequencing would optimize incompatibility assessment and cost-effectiveness of HLA typing, considering not only well-imputed second field typing(s) but also well-imputed eplets.

Predictive value of mixed antigen screen beads in pre-transplant assessment of HLA immunization in solid organ transplant recipients.

Pre-transplant serum screening of anti-HLA antibodies is recommended for solid organ transplantations. Many laboratories use the less expensive bead-based screening assay as the main technique and, if positive, turn to single-antigen beads (SAB). We studied the correlations between these two immunoassays. We re-analyzed the raw data of the two assays in 3030 first organ transplant recipients, explored with the two tests. We performed a ROC curve analysis of the screening ratio to predict a positive SAB assay. The AUC were 0.72 and 0.64 for class I and class II. The optimal thresholds of screening ratios were 3.28 (class I) and 2.11 (class II). Whatever the class, the negative predictive value was low, around 40%, with 36% of discordant sera, as defined by negative screening and positive SAB. Testing class I discordant sera on acid-treated SAB showed that 54% of antibodies reacted against denatured HLA molecules. However, these screening-negative sera may contain donor-specific antibodies in 13.9% and 28.7% of cases for class I and class II, respectively, involved in antibody-mediated rejection with the same frequency as non-discordant sera. Given the low predictive value of screening, both assays should be performed at least once on the same serum before transplantation.